The AI Came Later

Five weeks ago my job changed. The brief was to lead an AI transformation across a listed property group with five business units. Here is the honest report from month one: there was very little AI in it.

The reason is simple. Before a company can use AI, it has to be able to see itself. Most companies see themselves once a month, when the reports arrive. Everything that happens in between lives in phone calls and memory. You cannot train a model on memory. So the first job was not intelligence. It was sight.

Sight

That is why the first move went to property management, not sales. Every guard, property officer and manager got a reporting app on their phone. A defect, a visitor, a complaint, an incident: logged where it happens, time-stamped, visible upstairs the same day. Nothing about a guard tapping a phone is artificial intelligence. But a building that leaves no record of itself has nothing for a model to read. This was the raw material.

Once the raw material existed, the dashboards followed, one for each business unit. Moving averages, standard deviation flags, a simple forecast. People ask how many hours these save. That is the wrong measure. The right one is latency: the gap between something happening and a director knowing about it. That gap was a month. It is now a day. A company that sees itself daily can decide daily. That change is bigger than any single tool.

Trust

Seeing is the easy half. The hard half is getting people to feed the system, and that comes down to trust. A business unit manager hears "AI" and hears "headcount." The fear is not foolish; it is how these projects are usually sold. So the pitch had to start somewhere else. Look at your month-end. Count the days accounting takes from people who should be selling. The system is there to give those days back, not to take the people away. Until a manager believes that, the app stays uninstalled and the dashboard stays empty.

Who believed first tells you something. The youngest staff, given permission, built their own shortcuts within days, unprompted, because it made their own work lighter. Managers moved slower, not out of resistance but because nobody had ever shown them what was possible. Those are two different education problems, and one course does not cover both.

The chairman's questions

He did not ask how much we had saved. He asked two things. With the history loaded, can we decide faster and see trouble earlier? And could the method itself become a business? The first has a yes already. The second is a longer story.

Five weeks proves little, and it is worth saying so. Dashboards are not transformation. Plenty of companies build beautiful screens and change nothing behind them. The difficult work, contracts, receivables, customer enquiries, has not been touched. And property remains a business of relationships.

No dashboard closes a deal. But a company that sees itself every day manages every deal better. That was month one. The AI came later.